Glacier jump prediction and data sample enhancement management method

Abstract

The invention discloses a glacier jump prediction and data sample enhancement management method, particularly relates to the field of glacier remote sensing monitoring, is used for solving the problem of accurate identification and processing of surface moraine covering type glacier local albedo abnormal effect, and aims at solving the problem of accurate identification and processing of the glacier local albedo abnormal effect by comprehensively analyzing the coupling relationship between spectrum and temperature and the spatial sensitivity of temperature gradient. And accurate identification and processing of the moraine covering glacier local albedo abnormal effect are realized. According to the method, main causes of albedo abnormity can be effectively distinguished, and an abnormal area is corrected in a targeted manner by scientifically selecting an optical and thermal infrared remote sensing fusion mode or an optical and radar remote sensing fusion mode, so that the glacier identification precision and the jump prediction accuracy are remarkably improved. Through multi-level remote sensing data correction and feature extraction, a high-precision data foundation is laid for subsequent analysis; in the process of generating an enhanced sample and optimizing a prediction model, the reliability and response speed of glacier jump prediction are further improved.

Description

Technical Field

[0001] The present invention relates to the field of glacier remote sensing monitoring, and more specifically, to a method for glacier surge prediction and data sample enhancement management. Background Art

[0002] As an important water resource reserve on the earth, the dynamic changes of glaciers have a profound impact on the global climate, sea level changes and ecological environment. Therefore, the monitoring and analysis of glaciers has become an important topic of global concern. Remote sensing technology has become one of the main means of glacier monitoring due to its advantages of large range, continuity and high precision. Through multi-level processing and analysis of remote sensing images, it is possible to obtain information on the physical properties, dynamic changes and internal structure of glaciers, thereby providing a scientific basis for glacier surge prediction, disaster warning and water resource management. However, different types of glaciers, especially moraine-covered glaciers, have complex spectral reflectance and temperature distribution characteristics due to the uneven gravel and soil layers on their surfaces, which increases the difficulty of glacier type identification and surge prediction.

[0003] Existing technologies have significant deficiencies in processing remote sensing data of moraine-covered glaciers, mainly manifested in the lack of singleness and flexibility of analysis methods. First, traditional methods often focus on the analysis of single spectrum or temperature data, ignoring the complex interaction between the glacier surface spectrum and temperature, resulting in difficulty in obtaining accurate results when identifying and correcting local albedo anomalies. Secondly, existing technologies mostly rely on fixed data processing modes, and are usually unable to dynamically adjust according to specific glacier surface characteristics or environmental conditions. For example, for albedo anomalies caused by uneven thickness of the moraine layer, existing methods are difficult to effectively distinguish between its different causes and other types of anomalies, resulting in a high misjudgment rate. These deficiencies directly affect the accuracy of glacier type identification and the reliability of surge prediction, limiting the application effect of remote sensing technology in complex glacier monitoring.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiment of the present invention provides a method for glacier surge prediction and data sample enhancement management. First, through multi-level remote sensing data correction, the errors caused by sensor noise, atmospheric interference and terrain factors are successfully eliminated, and the real spectral information of the glacier surface is restored, which lays a high-precision data foundation for subsequent precise analysis. Combining multi-scale analysis and joint feature extraction, it is possible to deeply explore the complex spectral and texture characteristics of surface moraine-covered glaciers, ensure accurate classification of glacier areas in multi-dimensional feature space, and achieve accurate separation and clear spatial positioning of surface moraine-covered glaciers by generating binary mask maps. For the local albedo anomaly effect, by comprehensively analyzing the coupling relationship between spectrum and temperature and the spatial sensitivity of temperature gradient, it is possible to effectively distinguish the causes of anomalies, and by scientifically selecting optical and thermal infrared remote sensing fusion modes or optical and radar remote sensing fusion modes, it is ensured that each anomaly is optimally processed. This comprehensive evaluation method not only improves the accuracy of glacier identification and reduces the risk of misjudgment, but also enhances the adaptability to complex glacier environments, providing reliable technical guarantees for accurate glacier monitoring and disaster warning. In addition, by accurately extracting and reconstructing local features of the glacier surface and generating representative enhanced samples, the problem of insufficient data on real surge events is solved, so that the prediction model can more realistically reflect the actual changes of glaciers under complex conditions, and improve the sensitivity to signs of stress concentration or crack formation inside the glacier, so as to perceive potential surge risks in advance, provide more timely and effective warnings, and protect the safety of people and infrastructure in downstream areas. By continuously optimizing the accuracy and adaptability of the model, the glacier monitoring system can continue to learn and improve, make the most appropriate predictions and responses under different glacier types and changing conditions, provide solid technical support for the monitoring and early warning of glacier surges, and significantly improve the actual effect of glacier disaster prevention, so as to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] S1, after multi-level correction of remote sensing data, the multi-scale spectrum, texture and spectrum-temperature interaction characteristics of moraine-covered glaciers are captured through wavelet transform and joint feature extraction. Based on this, a classifier is trained and the moraine-covered glacier area is separated and separated;

[0008] S2, based on the coupling relationship between spectrum and temperature and the spatial sensitivity of temperature gradient, the dominant factor of local albedo anomaly is determined by calculating the comprehensive evaluation value and deciding the remote sensing fusion mode;

[0009] S3, based on the selected remote sensing fusion mode, extracts and reconstructs the local features of the glacier surface, generates new enhanced samples, and trains and optimizes the glacier surge prediction model by analyzing the spectral and temperature characteristics of the samples, for real-time monitoring and early warning.

[0010] In a preferred embodiment, step S1 includes the following contents:

[0011] S1.1, using wavelet transform multi-scale decomposition technology, the spectral information in remote sensing data is decomposed into multi-scale features, including details and trend parts at different scales;

[0012] S1.2, combining spectrum and texture information to perform joint feature extraction, extracting joint feature vectors by analyzing spectrum data and its gradient information;

[0013] S1.3, using a bidirectional conversion function to perform interactive analysis of spectrum and temperature data, and calculate the spectrum-temperature interactive characteristic value;

[0014] S1.4, based on the extracted joint feature vector and the spectrum-temperature interaction characteristic value as the classification feature information of the surface moraine-covered glacier, the classifier is trained through the machine learning algorithm to identify the surface moraine-covered glacier. After the training is completed, the classifier is applied to the actual remote sensing data to divide the multi-dimensional feature space of the glacier area into different categories, and identify the characteristic area where the surface moraine-covered glacier is located; by parsing the boundaries generated by the classifier, the pixel area of ​​the surface moraine-covered glacier is stripped and separated, and the corresponding binary mask image is generated.

[0015] In a preferred embodiment, step S2 includes the following contents:

[0016] The logic of obtaining the spectrum-temperature coupling index is as follows:

[0017] a1, based on the pixel area of ​​the moraine-covered glacier and the corresponding binary mask image, the spectral reflectance data and temperature data are normalized;

[0018] a2, calculate the difference field between the spectral reflectance value and the temperature value, reflecting the inconsistency of the spectrum and temperature at different locations on the glacier surface; formula: ;

[0019] : Spectrum-temperature difference value, indicating the difference between spectrum and temperature at position )’s absolute difference;

[0020] :wavelength Normalized spectral reflectance value under ;

[0021] : Normalized temperature value;

[0022] a3, in order to capture the coupling relationship between spectral reflectance and temperature change, the difference field is locally integrated to calculate the coupling strength around each pixel; formula: ;

[0023] : Pixels The local coupling strength of

[0024] : The coefficient that adjusts the effect of the difference field on the coupling strength;

[0025] : Local integration area, representing pixel points The surrounding space range;

[0026] a4, based on the local coupling intensity, the spectrum-temperature coupling index is obtained by calculating the coupling intensity distribution in the entire glacier area; formula: ;

[0027] : Spectral-temperature coupling index, which indicates the coupling strength between spectrum and temperature in the entire glacier area;

[0028] : The area of ​​the glacier pixel region, used to normalize the coupling strength.

[0029] In a preferred embodiment, the temperature gradient sensitivity coefficient acquisition logic is:

[0030] b1, first, calculate the gradient of the temperature data in the pixel area of ​​the moraine-covered glacier that has been obtained; formula: ;

[0031] : Pixels The intensity of the temperature gradient at

[0032] and :Temperature data in and The first-order partial derivative in the direction represents the rate of change of temperature in these two directions;

[0033] b2, after obtaining the temperature gradient intensity, conduct consistency analysis on its direction; formula: ;

[0034] : Pixels With adjacent pixels The cosine similarity of the temperature gradient direction between the two groups;

[0035] : Pixels and The dot product of the gradient vector between them;

[0036] and : The modulus of the gradient vector;

[0037] b3, by integrating the temperature gradient intensity and the gradient direction consistency, the local gradient sensitivity of each pixel is calculated; formula: ;

[0038] : Pixels The local gradient sensitivity of

[0039] : Temperature gradient intensity within the integration area;

[0040] : Gradient direction consistency within the integration region;

[0041] : Integration area, representing pixel points The surrounding space range;

[0042] b4, calculate the temperature gradient sensitivity coefficient by normalizing and integrating the local gradient sensitivity of the entire glacier area; formula: ;

[0043] : Temperature gradient sensitivity coefficient, which indicates the sensitivity of temperature changes in the entire glacier area;

[0044] : The area of ​​the glacier pixel region, used to normalize the gradient sensitivity.

[0045] In a preferred embodiment, c1, first calculates the weighted difference field between the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient, which is used to reflect the difference intensity between the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient at different pixel points; formula: ;

[0046] : Pixels The weighted difference value of

[0047] : Pixels The spectrum-temperature coupling index value at ;

[0048] : Pixels The temperature gradient sensitivity coefficient value at ;

[0049] and : Coefficient for adjusting the exponential weighted intensity;

[0050] c2, accumulates the absolute value of the calculated weighted difference field to quantify the total difference between the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient in the local area; formula: ;

[0051] : Pixels The local difference accumulation value of

[0052] : The absolute value of the weighted difference value in the local integration area;

[0053] : Integration area, representing pixel points The surrounding space range;

[0054] c3, converts the local difference accumulation value into a comprehensive evaluation value through nonlinear mapping; formula: ;

[0055] : Pixels The comprehensive evaluation value of

[0056] : nonlinear mapping coefficient;

[0057] c4, make mode selection decisions based on the calculated comprehensive evaluation value; formula: ;

[0058] : The selected analysis mode is the optical and thermal infrared remote sensing fusion mode or the optical and radar remote sensing fusion mode;

[0059] : Optical and thermal infrared remote sensing fusion mode;

[0060] : Optical and radar remote sensing fusion mode;

[0061] c5, based on the selected optical and thermal infrared remote sensing fusion mode or optical and radar remote sensing fusion mode, a more comprehensive analysis and correction of local albedo anomalies is performed.

[0062] In a preferred embodiment, step S3 includes the following contents:

[0063] Extract local features of the glacier surface from remote sensing data after model processing; combine the spectrum-temperature coupling index and temperature gradient sensitivity coefficient at different spatial scales using multi-scale analysis methods; generate high-dimensional feature vectors by exponential function processing of the temperature gradient sensitivity coefficient for subsequent sample enhancement;

[0064] Using a generative adversarial network, the reconstructed local feature vector is used as input to generate new glacier surge samples; the input high-dimensional feature vector is decoded into new sample data through a decoder to enhance the training set of the model;

[0065] Analyze the generated enhanced samples to identify the conditions that lead to glacier surges; use the combined characteristics of the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient, and set the identification threshold of glacier surges by exponentially weighting the characteristic values ​​in the samples to form a comprehensive evaluation value for surge prediction;

[0066] Using enhanced samples and identified surge condition thresholds, a glacier surge prediction model is trained;

[0067] After the model training and verification is completed, the remote sensing data of the glacier area is monitored in real time, and the predicted probability of glacier surge is calculated in real time based on the latest calculated spectral-temperature coupling index and temperature gradient sensitivity coefficient. The possibility of glacier surge is judged through nonlinear processing of real-time data. When the predicted probability reaches or exceeds the warning trigger threshold, the glacier surge warning mechanism is triggered and a warning signal is issued in a timely manner.

[0068] Technical effects and advantages of the glacier surge prediction and data sample enhancement management method of the present invention:

[0069] 1. The present invention effectively eliminates the errors caused by sensor noise, atmospheric interference and terrain factors through multi-level remote sensing data correction, restores the true spectral information of the glacier surface, and provides a high-precision data basis for subsequent analysis. Secondly, through multi-scale analysis and joint feature extraction, it is possible to deeply explore the complex spectral and texture characteristics of surface moraine-covered glaciers, ensuring the accurate classification of glacier areas in multi-dimensional feature space. This method can not only peel off and accurately separate the pixel areas of surface moraine-covered glaciers, but also generate binary mask images, providing clear spatial positioning for subsequent analysis and processing. Finally, the verification and adjustment of the classification results further improve the accuracy of recognition, ensuring that the classifier can maintain a high level of accuracy and stability when facing different glacier types and environmental conditions. This process provides reliable data support for glacier surge prediction, improves the overall effect of analysis and prediction, and has important application value.

[0070] 2. The present invention realizes the accurate identification and processing of the local albedo anomaly effects of surface moraine-covered glaciers by comprehensively analyzing the coupling relationship between spectrum and temperature and the spatial sensitivity of temperature gradient. In practical applications, this comprehensive evaluation method can effectively distinguish the main causes of albedo anomalies. Whether it is spectral anomalies caused by temperature changes or albedo fluctuations caused by terrain structure, targeted corrections can be made with the support of different modes. By scientifically selecting the optical and thermal infrared remote sensing fusion mode or the optical and radar remote sensing fusion mode, it is ensured that each albedo anomaly can be optimally processed, making the glacier surge prediction model more accurate and stable. This effect not only improves the accuracy of glacier identification, but also reduces the risk of misjudgment in practice, enhances the adaptability to complex glacier environments, and provides reliable technical guarantees for accurate glacier monitoring and disaster warning.

[0071] 3. The present invention generates representative enhanced samples by accurately extracting and reconstructing local features of the glacier surface, thereby providing a more reliable data basis for glacier surge prediction. This process effectively solves the problem of insufficient data on real surge events, allowing the prediction model to more realistically reflect the actual changes of glaciers under complex conditions. In practical applications, this method not only improves the sensitivity of the prediction model to signs of stress concentration or crack formation inside the glacier, but also can perceive potential surge risks in advance, provide more timely and effective warnings, and protect the safety of people and infrastructure in downstream areas. In addition, by continuously optimizing the accuracy and adaptability of the model, the glacier monitoring system can continue to learn and improve, and can make the most appropriate predictions and responses when facing different glacier types and changing conditions, providing solid technical support for the monitoring and early warning of glacier surges, and significantly improving the actual effect of glacier disaster prevention. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 It is a flow chart of the glacier surge prediction and data sample enhancement management method of the present invention. DETAILED DESCRIPTION

[0073] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0074] Example 1

[0075] Figure 1 The present invention provides a glacier surge prediction and data sample enhancement management method, including:

[0076] S1, after multi-level correction of remote sensing data, the multi-scale spectrum, texture and spectrum-temperature interaction characteristics of moraine-covered glaciers are captured through wavelet transform and joint feature extraction. Based on this, the classifier is trained and the moraine-covered glacier area is stripped and separated.

[0077] S2, based on the coupling relationship between spectrum and temperature and the spatial sensitivity of temperature gradient, determines the dominant factor of local albedo anomaly by calculating the comprehensive evaluation value and decides the remote sensing fusion mode.

[0078] S3, based on the selected remote sensing fusion mode, extracts and reconstructs the local features of the glacier surface, generates new enhanced samples, and trains and optimizes the glacier surge prediction model by analyzing the spectral and temperature characteristics of the samples, for real-time monitoring and early warning.

[0079] In remote sensing images, different types of glaciers show different spectral and textural characteristics, but due to the influence of various factors such as atmosphere, sensor noise and terrain, it becomes difficult to accurately identify the glacier type directly from the original data. Through the steps of radiation correction, geometric correction and atmospheric correction, the true spectral reflectance characteristics of the glacier surface are restored to create conditions for accurately distinguishing glacier types. Furthermore, through multi-scale analysis and joint feature extraction, deep feature mining and classification are carried out for the complex spectral, textural and temperature characteristics of moraine-covered glaciers. Finally, according to these classification results, various types of glacier areas are accurately separated, laying a solid foundation for subsequent analysis and application.

[0080] Step S1 includes the following contents:

[0081] The raw data of remote sensing images will be affected by the atmosphere, sensor noise and terrain factors during the acquisition process. Therefore, before using remote sensing data for glacier type identification, the data needs to be corrected at multiple levels. First, radiation correction is performed to eliminate the impact of factors such as sensor response inconsistency and sensor aging on the data. Then, geometric correction is performed to align the spatial distribution of the image with the geographic reference system to ensure that data from different bands can be accurately superimposed. Finally, atmospheric correction is implemented to calculate the scattering and absorption effects of the atmosphere on the spectrum through a model, thereby restoring the true spectral reflectance characteristics of the glacier surface. The above process is accurately described by formulas, and the processing results are corrected high-precision remote sensing data, which lays the foundation for the subsequent identification of glacier types.

[0082] S1.1, the surface reflectance characteristics of moraine-covered glaciers are complex and diverse, so it is necessary to perform multi-scale analysis on the spectral information of remote sensing data to capture the spectral characteristics of the glacier surface at different scales. Multi-scale decomposition techniques such as wavelet transform are used to decompose the spectral information in remote sensing data into multi-scale features, including details and trend parts at different scales, thereby revealing the spectral variation characteristics of moraine cover layers at micro and macro scales. By reconstructing multi-scale features, the feature expression ability of moraine cover layers is enhanced, ensuring that moraine-covered glacier areas can be accurately identified in subsequent processing.

[0083] S1.2, moraine-covered glaciers not only have similarities with the surrounding environment in terms of spectrum, but also show complex structural characteristics in terms of surface texture. Therefore, relying solely on spectral information may not accurately distinguish moraine-covered layers from other types of glaciers or terrain features. To solve this problem, spectral and texture information are combined for joint feature extraction. By analyzing spectral data and its gradient information, a joint feature vector is extracted. This joint feature vector can reflect the spectral and texture characteristics of moraine-covered layers, providing richer feature information for subsequent classification.

[0084] S1.3, moraine-covered glaciers have unique interactive characteristics in spectral and temperature distribution, that is, the physical characteristics of the moraine layer will have a significant impact on the spectral reflectance and temperature changes of the glacier surface. By analyzing the interactive relationship between spectrum and temperature, the characteristic pattern of the moraine cover layer can be revealed. In this step, the bidirectional conversion function is used to perform interactive analysis of spectral and temperature data, and the spectrum-temperature interaction characteristic value is calculated. This characteristic value can effectively reflect the spectral and temperature change characteristics of moraine-covered glaciers, and provides a key basis for the subsequent identification of glacier types. The spectrum and temperature data are combined through integral operation to generate the spectrum-temperature interaction characteristic value for classification.

[0085] S1.4, based on the joint feature vector extracted in the previous step and the spectrum-temperature interaction characteristic value, sufficient feature information is provided for the classification of surface moraine-covered glaciers. Using the classification feature information, the classifier is trained through a machine learning algorithm to accurately identify surface moraine-covered glaciers. Classifier training is based on a large number of labeled glacier data sets, and the classification performance is optimized through models such as support vector machines and random forests. After training, the classifier is applied to actual remote sensing data to divide the multidimensional feature space of the glacier area into different categories and identify the characteristic areas where the surface moraine-covered glaciers are located. By parsing the boundaries generated by the classifier, the pixel areas of the surface moraine-covered glaciers are stripped and separated, and the corresponding binary mask map is generated to achieve accurate separation of surface moraine-covered glaciers.

[0086] In this step, the formula describes in detail the input and output of the classification model, clarifies how to use the extracted features for classification, and ensures the accuracy of the classification results.

[0087] S1.5, the accuracy of the classification results directly affects the effectiveness of glacier surge prediction and other related analyses. Therefore, after identifying the moraine-covered glaciers, the results need to be verified and adjusted according to the actual observation data. By comparing the classification results with the actual glacier type data, possible misclassified areas are identified. For these areas, by combining contextual information such as terrain and climate for analysis, the classifier parameters are adjusted or the model is retrained to improve the accuracy of recognition. The formula processing process in this step describes in detail the calculation method of the adjustment amount to ensure the consistency of the classification results with the actual situation and provide a reliable data basis for subsequent analysis.

[0088] The present invention effectively eliminates the errors caused by sensor noise, atmospheric interference and terrain factors through multi-level remote sensing data correction, restores the true spectral information of the glacier surface, and provides a high-precision data basis for subsequent analysis. Secondly, through multi-scale analysis and joint feature extraction, the complex spectral and texture characteristics of surface moraine-covered glaciers can be deeply explored to ensure the accurate classification of glacier areas in multi-dimensional feature space. This method can not only peel off and accurately separate the pixel areas of surface moraine-covered glaciers, but also generate binary mask images to provide clear spatial positioning for subsequent analysis and processing. Finally, the verification and adjustment of the classification results further improve the accuracy of recognition, ensuring that the classifier can maintain a high level of accuracy and stability when facing different glacier types and environmental conditions. This process provides reliable data support for glacier surge prediction, improves the overall effect of analysis and prediction, and has important application value.

[0089] Due to the presence of the overburden, the surface albedo of moraine-covered glaciers may vary abnormally in different areas and is difficult to unify. This local albedo anomaly increases the complexity of glacier identification, especially when the thickness of the overburden is uneven. In pure glaciers and mountain-shadowed glaciers, the albedo is usually relatively stable or evenly reduced due to the shadow.

[0090] For the pixel areas of moraine-covered glaciers that have been obtained, the analysis of the "local albedo anomaly effect" can reveal the inhomogeneity and abnormal changes in the reflectivity of the glacier surface. Especially under the coverage of moraine layers of different thicknesses, the local anomaly of albedo may indicate the instability of the glacier's internal structure or potential precursors of surge. The analysis of this effect can help identify abnormal reflective areas on the glacier surface, which may be associated with stress concentration, crack formation and heat accumulation of the glacier.

[0091] In terms of glacier surge prediction, the local albedo anomaly effect can serve as an important early warning indicator. By monitoring and analyzing the changes in these abnormal areas, the possibility and time of glacier surge can be predicted. For data sample enhancement, the identification and analysis of the local albedo anomaly effect can generate more representative sample data, enhance the model's ability to identify glacier surges, and improve prediction accuracy. These samples can better reflect the actual glacier surge process, optimize model training, and improve the prediction effect of glacier surges.

[0092] As an important method for dealing with the effects of local albedo anomalies, multi-source data fusion can comprehensively utilize the advantages of different remote sensing data sources to comprehensively analyze and correct the spectral, structural and temperature characteristics of moraine-covered glaciers. Radar remote sensing (SAR) is highly sensitive to terrain undulations and surface cover, and can effectively penetrate the moraine layer to provide information on the structure below the glacier surface, thereby helping to identify areas of albedo anomalies caused by changes in the thickness of the moraine layer. Thermal infrared remote sensing is good at capturing temperature differences on the glacier surface and can reveal albedo anomalies caused by temperature changes. The fusion of these data can form a comprehensive perspective to more accurately locate and analyze local albedo anomalies and improve the monitoring and prediction capabilities of moraine-covered glaciers.

[0093] In the process of multi-source data fusion, two analysis modes, optical and radar remote sensing fusion mode and optical and thermal infrared remote sensing fusion mode, are used to deal with different causes of albedo anomalies. In the optical and radar remote sensing fusion mode, the spectral information of optical images and the penetration ability of radar images are mainly combined, and the structural change areas under the moraine layer that may cause albedo anomalies are accurately identified through multi-dimensional feature joint analysis. This mode is particularly suitable for scenes where the albedo anomaly is highly correlated with the physical properties of the moraine layer, such as thickness and density, and can effectively correct the albedo anomaly caused by the surface structure.

[0094] The optical and thermal infrared remote sensing fusion mode focuses on analyzing the correlation between spectral reflectance and temperature changes. Combining the fine spectral information of optical images and the temperature distribution of thermal infrared images, it can identify areas of albedo anomaly caused by abnormal temperature of the moraine layer. This mode is suitable for situations where albedo anomaly is closely related to changes in glacier surface temperature. By analyzing the spectrum-temperature interaction characteristics, it can effectively correct albedo anomalies caused by temperature fluctuations.

[0095] The application of these two analysis modes in multi-source data fusion can select the most suitable processing method for the causes of different local albedo anomaly effects, thereby enhancing the ability to accurately identify and predict the surge of moraine-covered glaciers. This processing method can not only improve the utilization efficiency of remote sensing data, but also optimize the quality and reliability of sample data through comprehensive analysis of different modes, providing stronger support for glacier surge prediction models.

[0096] Step S2 includes the following contents:

[0097] The reflectance characteristics and temperature distribution of the glacier surface have a complex interactive relationship under different environmental conditions, especially in moraine-covered glaciers, where this relationship is more significant. The presence of the moraine layer makes the temperature change on the glacier surface directly affect its spectral reflectance characteristics, thereby causing albedo anomalies. In order to more accurately identify and correct these anomalies, the spectral-temperature coupling index was proposed as a key indicator to measure the coupling relationship between the spectral characteristics of the glacier surface and temperature changes.

[0098] The calculation of the spectral-temperature coupling index is based on a reasonable assumption: there is a certain correlation between the spectral reflectance characteristics of the glacier surface and its temperature changes, especially in areas with thick moraine cover, where this coupling effect is more significant. By quantifying this coupling relationship, the specific impact of temperature fluctuations on spectral reflectance can be revealed, thereby providing more accurate anomaly detection and correction methods.

[0099] In remote sensing data processing, traditional spectral or temperature analysis can only capture a certain characteristic of the glacier surface separately, but cannot reflect the complex interaction between the two. The spectral-temperature coupling index provides a new analytical perspective by integrating spectral and temperature data, which can more comprehensively understand the dynamic changes of moraine-covered glaciers. This index not only provides more detailed support for glacier surge prediction, but also provides a theoretical basis for the fusion of multi-source remote sensing data, ensuring that the analysis results under complex environmental conditions are more reliable and accurate.

[0100] The introduction of the spectrum-temperature coupling index not only helps to cope with the challenge of albedo anomaly effect, but also provides reasonable scientific support for the accurate combination of temperature and spectrum data in glacier environmental monitoring. The calculation method of this index has shown wide applicability and high efficiency in practical applications, and can provide strong support for the in-depth analysis of complex glacier systems.

[0101] The logic of obtaining the spectrum-temperature coupling index is as follows:

[0102] a1, based on the pixel area of ​​the moraine-covered glacier and the corresponding binary mask map, the spectral reflectance data and temperature data are normalized. The purpose of normalization is to eliminate the scale difference between different data sources and ensure that the two data can be coupled and analyzed at the same scale.

[0103] formula:

[0104] ;

[0105] ;

[0106] :wavelength Normalized spectral reflectance value under .

[0107] : Original spectral reflectance value.

[0108] and : Minimum and maximum values ​​of spectral reflectance.

[0109] : Normalized temperature value.

[0110] : Raw temperature data.

[0111] and : Minimum and maximum values ​​of temperature data.

[0112] a2, calculate the difference field between the spectral reflectance and the temperature value, reflecting the inconsistency of the spectrum and temperature at different locations on the glacier surface. The calculation of the difference field can reveal local mismatch areas between the spectrum and temperature, which may be potential sources of albedo anomalies.

[0113] formula: ;

[0114] : Spectrum-temperature difference value, indicating the difference between spectrum and temperature at position ) absolute difference.

[0115] a3,In order to capture the coupling relationship between spectral reflectance and temperature change, the difference field is locally integrated and the coupling strength around each pixel is calculated. The local coupling strength reflects the correlation between spectral reflectance and temperature in a specific area and is an important component of the spectrum-temperature coupling index.

[0116] formula: ;

[0117] : Pixels The local coupling strength.

[0118] : Coefficient that adjusts the effect of the difference field on the coupling strength.

[0119] : Local integration area, representing pixel points The surrounding space.

[0120] a4, based on the local coupling intensity, the spectrum-temperature coupling index is obtained by calculating the coupling intensity distribution in the entire glacier area. This index is used to quantify the overall coupling degree between spectrum and temperature in the glacier area and provide support for mode selection. When the coupling index is high, it indicates that the spectrum and temperature have a strong correlation, and it is suitable to select the optical and thermal infrared remote sensing fusion mode for albedo anomaly analysis.

[0121] formula: ;

[0122] : Spectrum-temperature coupling index, which indicates the coupling strength between spectrum and temperature in the entire glacier area.

[0123] : The area of ​​the glacier pixel region, used to normalize the coupling strength.

[0124] The application of the spectral-temperature coupling index in glacier remote sensing analysis has significantly improved the ability to handle local albedo anomaly effects. First, the index can accurately capture the coupling characteristics of spectral reflectance and temperature changes on the glacier surface, ensuring accurate identification and correction of albedo anomaly areas. Secondly, through the quantitative analysis of the coupling index, the optimal data fusion mode can be scientifically selected to fully utilize the advantages of optical, radar and thermal infrared remote sensing data to enhance the interpretation of remote sensing data. In addition, the spectral-temperature coupling index also plays an important role in glacier surge prediction and sample enhancement. By providing high-precision decision support, it significantly improves the prediction performance and generalization ability of the model. It provides strong technical support for remote sensing data processing in complex glacial environments and has broad application prospects and practical value.

[0125] Temperature changes on the glacier surface have an important impact on its reflectivity and overall stability. Especially in moraine-covered glaciers, changes in temperature gradients are often one of the key causes of albedo anomalies. Due to the coverage of the moraine layer, the temperature distribution on the glacier surface becomes complex and uneven, resulting in temperature gradients in local areas that may cause significant reflectivity fluctuations. In order to accurately identify these anomalies and make effective corrections, the temperature gradient sensitivity coefficient has emerged as a quantitative indicator.

[0126] The temperature gradient sensitivity coefficient aims to capture the influence of temperature gradient on spectral reflectance characteristics by quantifying the spatial sensitivity of glacier surface temperature changes. This coefficient reveals the complex relationship between temperature change and spectral anomaly by accurately calculating the intensity of temperature gradient and its directional consistency, providing a scientific basis for the detection and analysis of albedo anomalies. In particular, in remote sensing data processing and fusion mode selection, this coefficient provides a reasonable basis for identifying and responding to changes in glacier behavior under complex temperature conditions.

[0127] The introduction of temperature gradient sensitivity coefficient provides a necessary tool for further understanding of physical phenomena on the glacier surface and in-depth analysis of remote sensing data. It not only provides rational support for dealing with local albedo anomaly effects, but also provides a scientific basis for its application in actual glacier monitoring and prediction.

[0128] Temperature gradient sensitivity coefficient acquisition logic:

[0129] b1. First, the temperature data in the pixel area of ​​the moraine-covered glacier is gradient calculated. The gradient calculation aims to capture the spatial change rate of the glacier surface temperature and reflect the local change characteristics of the temperature at different locations. The larger the absolute value of the gradient, the more drastic the temperature change in the area.

[0130] formula: ;

[0131] : Pixels The strength of the temperature gradient at .

[0132] and :Temperature data in and The first-order partial derivative in the direction represents the rate of change of temperature in these two directions.

[0133] b2. After obtaining the temperature gradient intensity, perform consistency analysis on its direction. The consistency of the temperature gradient direction reflects the spatial characteristics of the temperature change in the region, especially whether the temperature change has obvious directionality. By calculating the cosine similarity of the gradient direction, analyze whether the temperature change directions between adjacent pixels are consistent.

[0134] formula: ;

[0135] : Pixels With adjacent pixels The cosine similarity of the temperature gradient direction between the two images.

[0136] : Pixels and The dot product of the gradient vectors.

[0137] and : The modulus of the gradient vector.

[0138] b3, by integrating the temperature gradient intensity and the gradient direction consistency, the local gradient sensitivity of each pixel is calculated. The local gradient sensitivity indicates the sensitivity of the temperature change around the pixel, taking into account both the magnitude of the temperature change and the direction of the change.

[0139] formula: ;

[0140] : Pixels The local gradient sensitivity.

[0141] : Temperature gradient intensity within the integration region.

[0142] : Gradient direction consistency within the integration region.

[0143] : Integration area, representing pixel points The surrounding space.

[0144] b4, the temperature gradient sensitivity coefficient is calculated by normalizing and integrating the local gradient sensitivity of the entire glacier area. This coefficient is used to quantify the overall sensitivity of temperature changes in the glacier area and provide support for model selection. When the temperature gradient sensitivity coefficient is high, it indicates that temperature changes have a greater impact on the glacier albedo anomaly, and it is suitable to select the optical and thermal infrared remote sensing fusion mode for analysis.

[0145] formula: ;

[0146] : Temperature gradient sensitivity coefficient, which indicates the sensitivity of temperature changes in the entire glacier area.

[0147] : The area of ​​the glacier pixel region, used to normalize the gradient sensitivity.

[0148] The application of temperature gradient sensitivity coefficient in glacier remote sensing analysis has significantly enhanced the understanding of the impact of temperature change on the spectral characteristics of the glacier surface. First, the coefficient can accurately capture the spatial distribution characteristics of the temperature gradient on the glacier surface, especially under the moraine cover layer. The impact of this gradient change on the albedo anomaly is particularly important. Secondly, the temperature gradient sensitivity coefficient provides a quantitative basis for the selection of fusion mode of remote sensing data, making the analysis process more accurate and reliable, especially in the optimal combination between optical and thermal infrared remote sensing, which can effectively improve the efficiency and accuracy of albedo anomaly correction. In addition, through the application of temperature gradient sensitivity coefficient, the prediction model of glacier surge can be further optimized, the representativeness of sample data is enhanced, and the robustness and generalization ability of prediction are improved. It provides stronger technical support for glacier environmental monitoring and has broad application prospects and value.

[0149] In order to solve the local albedo anomaly effect, a comprehensive evaluation value is generated by combining the spectral-temperature coupling index (STC) and the temperature gradient sensitivity coefficient (TGS) to indicate whether processing mode 1 (optical and radar remote sensing fusion mode) or processing mode 2 (optical and thermal infrared remote sensing fusion mode) should be used.

[0150] c1, first calculate the weighted difference field between the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient. This difference field is used to reflect the difference intensity between the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient at different pixel points. By exponentially processing the two, the nonlinear characteristics of the index difference are enhanced.

[0151] formula: ;

[0152] : Pixels The weighted difference value of .

[0153] : Pixels The spectrum-temperature coupling index value at .

[0154] : Pixels The temperature gradient sensitivity coefficient value at .

[0155] and : Adjust the coefficient of the exponential weighted intensity to control the influence of the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient in the difference field.

[0156] c2, accumulate the absolute value of the calculated weighted difference field to quantify the total difference between the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient in the local area. This step spatially accumulates the absolute value of the weighted difference field of each pixel point to capture its difference characteristics in the local area and further integrate the influence of the two.

[0157] formula: ;

[0158] : Pixels The local difference accumulation value.

[0159] : The absolute value of the weighted difference within the local integration region.

[0160] : Integration area, representing pixel points The surrounding space.

[0161] c3, converts the local difference accumulation value into a comprehensive evaluation value through nonlinear mapping. This mapping function is used to map the accumulated difference value to the indication value range of the mode selection, thereby obtaining the final evaluation result. By using the hyperbolic tangent function, it is ensured that the nonlinear characteristics in the mapping process are fully expressed, so that the change of the evaluation value is more in line with the complexity in the actual scene.

[0162] formula: ;

[0163] : Pixels A comprehensive evaluation value of , which is used to indicate the processing mode used.

[0164] : Nonlinear mapping coefficient, used to adjust the impact of cumulative differences on the evaluation value.

[0165] c4, make a mode selection decision based on the calculated comprehensive evaluation value. The positive or negative evaluation value determines which processing mode is selected. When the evaluation value is positive, it means that the spectrum-temperature coupling relationship is dominant, and it is suitable to select mode 2 (optical and thermal infrared remote sensing fusion mode); when the evaluation value is negative, the temperature gradient change is dominant, and it is suitable to select mode 1 (optical and radar remote sensing fusion mode).

[0166] formula: ;

[0167] :The selected analysis mode is the optical and thermal infrared remote sensing fusion mode (Optical-TIRFusion) or the optical and radar remote sensing fusion mode (Optical-SARFusion).

[0168] : Optical and thermal infrared remote sensing fusion model.

[0169] : Optical and radar remote sensing fusion mode.

[0170] c5, based on the selected optical and thermal infrared remote sensing fusion mode or optical and radar remote sensing fusion mode, a more comprehensive analysis and correction of local albedo anomalies is performed.

[0171] After selecting the optical and thermal infrared remote sensing fusion mode, the spectral characteristics of each band on the glacier surface are first analyzed through optical images to identify abnormal reflection areas. Then, thermal infrared images are used to capture the temperature distribution and temperature gradient changes on the glacier surface, and the temperature data is compared and analyzed with the spectral characteristics. By analyzing the spatial relationship between the temperature anomaly area and the spectral reflectance anomaly area, it is determined which albedo anomalies are caused by temperature changes. Finally, the temperature information is used to correct the spectral reflectance, especially for the high or low reflection areas caused by temperature anomalies, and appropriate adjustments are made to make the spectral reflectance data more accurately reflect the actual situation on the glacier surface.

[0172] After selecting the optical and radar remote sensing fusion mode, the spectral characteristics of the glacier surface are first extracted through optical images to identify areas with abnormal albedo. Then, radar remote sensing data (SAR) is used to analyze the structural characteristics and topographic changes below the glacier surface, with particular attention to the thickness and density changes of the moraine layer. By combining optical data with radar data, how these structural characteristics affect the spectral reflectance characteristics of the surface is analyzed. In particular, for albedo anomalies caused by undulating terrain and uneven thickness of the moraine layer, the depth information provided by the radar data is used for correction. Finally, the structural characteristics obtained from the radar remote sensing data are fused with the spectral reflectance information to correct the abnormal albedo areas in the optical image, so that the spectral data can more accurately reflect the true surface conditions of the glacier.

[0173] The present invention realizes the accurate identification and processing of the local albedo anomaly effects of surface moraine-covered glaciers by comprehensively analyzing the coupling relationship between spectrum and temperature and the spatial sensitivity of temperature gradient. In practical applications, this comprehensive evaluation method can effectively distinguish the main causes of albedo anomalies. Whether it is spectral anomalies caused by temperature changes or albedo fluctuations caused by terrain structure, targeted corrections can be made with the support of different modes. By scientifically selecting the optical and thermal infrared remote sensing fusion mode or the optical and radar remote sensing fusion mode, it is ensured that each albedo anomaly can be optimally processed, making the glacier surge prediction model more accurate and stable. This effect not only improves the accuracy of glacier identification, but also reduces the risk of misjudgment in practice, enhances the adaptability to complex glacier environments, and provides reliable technical guarantees for accurate glacier monitoring and disaster warning.

[0174] Step S3 includes the following contents:

[0175] After pattern processing, the local features of the glacier surface are first extracted and reconstructed based on the obtained spectrum-temperature coupling index and temperature gradient sensitivity coefficient. The extracted local features reflect the spectral and temperature variation characteristics of the glacier at different spatial scales. The spectrum-temperature coupling index and temperature gradient sensitivity coefficient at different scales are combined using a multi-scale analysis method, and the influence of the temperature gradient is adjusted through an exponential function to generate a high-dimensional feature vector.

[0176] The local features of the glacier surface are extracted from the remote sensing data after model processing. The spectrum-temperature coupling index and the temperature gradient sensitivity coefficient at different spatial scales are combined using a multi-scale analysis method. By processing the temperature gradient sensitivity coefficient with an exponential function, a comprehensive high-dimensional feature vector is generated for subsequent sample enhancement.

[0177] formula: ;

[0178] : Pixels The multi-scale local feature vector of .

[0179] : No. Spectral-temperature coupling index at this scale.

[0180] : No. The temperature gradient sensitivity coefficient at this scale.

[0181] : The number of scales for multiscale analysis.

[0182] : Adjustment coefficient for adjusting the influence of temperature gradient sensitivity coefficient.

[0183] Then, the reconstructed high-dimensional feature vector is used as input to generate new sample data using the generative adversarial network (GAN) or variational autoencoder (VAE) technology. The sample generation process includes spatial accumulation and nonlinear transformation of feature vectors to ensure that the generated samples can retain the spectrum-temperature coupling characteristics and temperature gradient change characteristics. The decoder is used to decode the input high-dimensional feature vector into new samples, which are used to enhance the data set of the glacier surge prediction model, thereby improving the recognition ability of the model.

[0184] Using a generative adversarial network (GAN) or a variational autoencoder (VAE), the reconstructed local feature vector is used as input to generate new glacier surge samples. These samples retain the glacier surface spectrum-temperature coupling characteristics and temperature gradient change characteristics. The input high-dimensional feature vector is decoded into new sample data through a decoder to enhance the model's training set.

[0185] formula: ;

[0186] : The new sample generated.

[0187] : Decoder function, used to decode the feature vector into sample data.

[0188] : The input noise vector for a generative adversarial network or variational autoencoder.

[0189] : Integration area, which represents the local spatial feature accumulation area when the sample is generated.

[0190] Based on the generated enhanced samples, the key conditions that may cause glacier surges are identified through nonlinear mapping and integral operations. The identification threshold of glacier surges is set by using the combined characteristics of the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient. The threshold calculation process includes exponential weighting of the characteristic values ​​in the samples to form a comprehensive evaluation value for surge identification. This evaluation value is used to train and verify the glacier surge prediction model.

[0191] The generated enhanced samples are analyzed to identify the conditions that may lead to glacier surges. The combined characteristics of the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient are used to set the identification threshold of glacier surges through exponential weighting of the characteristic values ​​in the samples, forming a comprehensive evaluation value for surge prediction.

[0192] formula: ;

[0193] : Identification threshold of glacier surge.

[0194] and : Coefficient for adjusting the influence of spectrum-temperature coupling index and temperature gradient sensitivity coefficient.

[0195] : Coefficient for adjusting the influence of local eigenvectors.

[0196] : The spatial scope of the recognition of jump conditions.

[0197] During the training process of the glacier surge prediction model, the model is optimized using logistic regression or deep neural network combined with enhanced samples and identified surge conditions. The model training takes into account the nonlinear effects of the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient, and the output of the model is the predicted probability of glacier surge. The generalization ability and prediction accuracy of the model are verified through cross-validation or Monte Carlo simulation to ensure the reliability of the model in practical applications.

[0198] The glacier surge prediction model is trained using enhanced samples and identified surge condition thresholds. The model training should fully consider the nonlinear effects of the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient to ensure that the model has high predictive ability for various possible surge conditions. The generalization ability and prediction accuracy of the model are verified by cross-validation or Monte Carlo simulation.

[0199] formula: ;

[0200] : Prediction probability of glacier surge.

[0201] , and : Weight parameters in model training.

[0202] , and : Input variables used for prediction model.

[0203] After the model training and verification is completed, the remote sensing data of the glacier area is monitored in real time, and the predicted probability of glacier surge is calculated in real time based on the latest calculated spectrum-temperature coupling index and temperature gradient sensitivity coefficient. The possibility of glacier surge is judged by nonlinear processing of real-time data. When the predicted probability reaches or exceeds the warning trigger threshold, the glacier surge warning mechanism is triggered, and a warning signal is issued in time to prevent potential glacier surge risks.

[0204] formula: ;

[0205] : Jump warning status.

[0206] : Warning trigger threshold, set according to the actual application scenario.

[0207] The present invention generates representative enhanced samples by accurately extracting and reconstructing local features of the glacier surface, thereby providing a more reliable data basis for glacier surge prediction. This process effectively solves the problem of insufficient data on real surge events, allowing the prediction model to more realistically reflect the actual changes of glaciers under complex conditions. In practical applications, this method not only improves the sensitivity of the prediction model to signs of stress concentration or crack formation inside the glacier, but also can perceive potential surge risks in advance, provide more timely and effective warnings, and protect the safety of people and infrastructure in downstream areas. In addition, by continuously optimizing the accuracy and adaptability of the model, the glacier monitoring system can continue to learn and improve, and can make the most appropriate predictions and responses when facing different glacier types and changing conditions, providing solid technical support for the monitoring and early warning of glacier surges, and significantly improving the actual effect of glacier disaster prevention.

[0208] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0209] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0210] It should be noted that, in this article, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including a..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0211] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for predicting glacier surges and enhancing data samples, characterized in that: Includes steps: S1, after multi-level correction of remote sensing data, the multi-scale spectrum, texture and spectrum-temperature interaction characteristics of moraine-covered glaciers are captured through wavelet transform and joint feature extraction. Based on this, a classifier is trained and the moraine-covered glacier area is separated and separated; S2, based on the coupling relationship between spectrum and temperature and the spatial sensitivity of temperature gradient, the dominant factor of local albedo anomaly is determined by calculating the comprehensive evaluation value and deciding the remote sensing fusion mode; S3, based on the selected remote sensing fusion mode, extracts and reconstructs the local features of the glacier surface, generates new enhanced samples, and trains and optimizes the glacier surge prediction model by analyzing the spectral and temperature characteristics of the samples, for real-time monitoring and early warning.

2. The method for glacier surge prediction and data sample enhancement management according to claim 1, characterized in that: Step S1 includes the following contents: S1.1, using wavelet transform multi-scale decomposition technology, the spectral information in remote sensing data is decomposed into multi-scale features, including details and trend parts at different scales; S1.2, combining spectrum and texture information to perform joint feature extraction, extracting joint feature vectors by analyzing spectrum data and its gradient information; S1.3, using a bidirectional conversion function to perform interactive analysis of spectrum and temperature data, and calculate the spectrum-temperature interactive characteristic value; S1.4, based on the extracted joint feature vector and the spectrum-temperature interaction characteristic value as the classification feature information of the surface moraine-covered glacier, the classifier is trained through the machine learning algorithm to identify the surface moraine-covered glacier. After the training is completed, the classifier is applied to the actual remote sensing data to divide the multi-dimensional feature space of the glacier area into different categories, and identify the characteristic area where the surface moraine-covered glacier is located; by parsing the boundaries generated by the classifier, the pixel area of ​​the surface moraine-covered glacier is stripped and separated, and the corresponding binary mask image is generated.

3. The method for glacier surge prediction and data sample enhancement management according to claim 2 is characterized in that: Step S2 includes the following contents: The logic of obtaining the spectrum-temperature coupling index is as follows: a1, based on the pixel area of ​​the moraine-covered glacier and the corresponding binary mask image, the spectral reflectance data and temperature data are normalized; a2, calculate the difference field between the spectral reflectance value and the temperature value, reflecting the inconsistency of the spectrum and temperature at different locations on the glacier surface; formula: ; : Spectrum-temperature difference value, indicating the difference between spectrum and temperature at position )’s absolute difference; :wavelength Normalized spectral reflectance value under ; : Normalized temperature value; a3, in order to capture the coupling relationship between spectral reflectance and temperature change, the difference field is locally integrated to calculate the coupling strength around each pixel; formula: ; : Pixels The local coupling strength of : The coefficient that adjusts the effect of the difference field on the coupling strength; : Local integration area, representing pixel points The surrounding space range; a4, based on the local coupling intensity, the spectrum-temperature coupling index is obtained by calculating the coupling intensity distribution in the entire glacier area; formula: ; : Spectral-temperature coupling index, which indicates the coupling strength between spectrum and temperature in the entire glacier area; : The area of ​​the glacier pixel region, used to normalize the coupling strength.

4. The method for glacier surge prediction and data sample enhancement management according to claim 3 is characterized by: Temperature gradient sensitivity coefficient acquisition logic: b1, first, calculate the gradient of the temperature data in the pixel area of ​​the moraine-covered glacier that has been obtained; formula: ; : Pixels The intensity of the temperature gradient at and :Temperature data in and The first-order partial derivative in the direction represents the rate of change of temperature in these two directions; b2, after obtaining the temperature gradient intensity, conduct consistency analysis on its direction; formula: ； : Pixels With adjacent pixels The cosine similarity of the temperature gradient direction between the two groups; : Pixels and The dot product of the gradient vector between them; and : The modulus of the gradient vector; b3, by integrating the temperature gradient intensity and the gradient direction consistency, the local gradient sensitivity of each pixel is calculated; formula: ; : Pixels The local gradient sensitivity of : Temperature gradient intensity within the integration area; : Gradient direction consistency within the integration region; : Integration area, representing pixel points The surrounding space range; b4, calculate the temperature gradient sensitivity coefficient by normalizing and integrating the local gradient sensitivity of the entire glacier area; formula: ; : Temperature gradient sensitivity coefficient, which indicates the sensitivity of temperature changes in the entire glacier area; : The area of ​​the glacier pixel region, used to normalize the gradient sensitivity.

5. The method for glacier surge prediction and data sample enhancement management according to claim 4 is characterized in that: c1, first calculate the weighted difference field between the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient, which is used to reflect the difference intensity between the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient at different pixel points; formula: ； : Pixels The weighted difference value of : Pixels The spectrum-temperature coupling index value at ; : Pixels The temperature gradient sensitivity coefficient value at ; and : Coefficient for adjusting the exponential weighted intensity; c2, accumulates the absolute value of the calculated weighted difference field to quantify the total difference between the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient in the local area; formula: ; : Pixels The local difference accumulation value of : The absolute value of the weighted difference value in the local integration area; : Integration area, representing pixel points The surrounding space range; c3, converts the local difference accumulation value into a comprehensive evaluation value through nonlinear mapping; formula: ； : Pixels The comprehensive evaluation value of : nonlinear mapping coefficient; c4, make mode selection decisions based on the calculated comprehensive evaluation value; formula: ； : The selected analysis mode is the optical and thermal infrared remote sensing fusion mode or the optical and radar remote sensing fusion mode; : Optical and thermal infrared remote sensing fusion mode; : Optical and radar remote sensing fusion mode; c5, based on the selected optical and thermal infrared remote sensing fusion mode or optical and radar remote sensing fusion mode, a more comprehensive analysis and correction of local albedo anomalies is performed.

6. The method for glacier surge prediction and data sample enhancement management according to claim 5 is characterized in that: Step S3 includes the following contents: Extract local features of the glacier surface from remote sensing data processed by models; The multi-scale analysis method is used to combine the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient at different spatial scales. The high-dimensional feature vector is generated by exponential function processing of the temperature gradient sensitivity coefficient for subsequent sample enhancement. Using a generative adversarial network, the reconstructed local feature vector is used as input to generate new glacier surge samples; the input high-dimensional feature vector is decoded into new sample data through a decoder to enhance the training set of the model; Analyze the generated enhanced samples to identify the conditions that lead to glacier surges; use the combined characteristics of the spectrum-temperature coupling index and the temperature gradient sensitivity coefficient, and set the identification threshold of glacier surges by exponentially weighting the characteristic values ​​in the samples to form a comprehensive evaluation value for surge prediction; Using enhanced samples and identified surge condition thresholds, a glacier surge prediction model is trained; After the model training and verification is completed, the remote sensing data of the glacier area is monitored in real time, and the predicted probability of glacier surge is calculated in real time based on the latest calculated spectral-temperature coupling index and temperature gradient sensitivity coefficient. The possibility of glacier surge is judged through nonlinear processing of real-time data. When the predicted probability reaches or exceeds the warning trigger threshold, the glacier surge warning mechanism is triggered and a warning signal is issued in a timely manner.

Images